Improving the Pulse Stability of Mode-Locked Diode Lasers
Sep 5, 2023
PLENARY SESSION II
Chair: James Coleman, Electrical and Computer Engineering and MNTL
CNST Nanotechnology Workshop 2013
Right, good morning. I, as most of you know, work on lasers and one of the things I'm particularly
interested in is mode-locked lasers. So, let's go, whoops, how do I move? Right, I'm going
to show a slide on some of the motivation and I'd like to give a brief introduction
to mode-locking lasers because I don't think most people know how they work. Then I'm going
to describe the work that we've done on mode-locked lasers, looking at 40 gigahertz mode-locked
lasers and the effects that we've been able to, looking at the structure, the geometry
of the laser and the epilayer structure changing that and then a quick summary. And I'm between
you and coffee so I'll be as quick as I can. Right, mode-locked lasers are basically used
in applications where you're looking for short pulses, particularly at high frequencies with
the semiconductor lasers. Semiconductor lasers, mode-lock, most easily it tends to about 100
gigahertz and that's related to the length of the cavity, that's the pulse round-trip
time. Bigger lasers give out more power and operate at much lower frequencies down to
the megahertz range. But they could be useful for the next generation of TDM at greater
than 100 gigabits, all optical recovery, clock acquisition, ultra-fast signal processing,
high-speed sampling where you put in electro-optic sampling, where you put in an optical pulse
and look at the electrical response. Microwave millimetre measurements, fibre radio transmit
high-frequency signal using optical fibre and mode-locked pulses to small transmitters
that are local transmitters, low power, analogue to digital conversion, high-speed logic. Right,
how do we make a laser mode-lock? Well, any laser that there is has several modes and
the intensity, the output intensity, is just a sum of the square of the electric field
of these modes as these modes add up. Now we have here the amplitude of the electric
field and this is the mode frequency and then there's this term here which is the phase.
That phase normally is random between modes. And if we look at it, you can see here this
is some modes and this is time along the bottom axis and you can see there's a random relationship
which changes with time between the different modes. And the output you get is Cw with some
noise on top of it as the phase relationship between the modes change, the way that this
adds up changes slightly so you get some noise on top of the Cw. Now if you can get into
this situation where at one point in time all the modes are in phase, these will all
add up and give you a high intensity pulse. What happens when they're out of phase, if
it's properly locked, they add up to zero, all the modes will add up to zero and all
the intensity of light is in this ultra-short pulse. How do we do that? We use a saturable
absorber. We put a saturable absorber in the cavity and this is called passive mode locking.
There are other forms but I'm not going into them. Basically with a saturable absorber,
the absorption decreases as the light intensity increases. You get a graph something like
this. At low intensity you have high absorption which as you increase the intensity it decreases.
Quantum wells will work very well in doing this because if you've got light at the band
gap energy, once you've filled up the states and the bottom of the wells, that well will
no longer absorb. So a short length of quantum well acts well as a saturable absorber. The
type of lasers I'm going to talk about is quantum well material. First of all I'm going
to talk about this structure which is a traditional structure. The lasers are operating at 1.55
microns so we're using slightly strained aluminium gallium indium arsenide wells which are grown
on an indium phosphide substrate. At the bottom here we have the indium phosphide substrate
and a cladding layer which is doped n-type. On the top we grow p-indium phosphide and
on top of that there will be a contact layer which is indium gallium arsenide which is
highly p-doped. In the middle in between we have the waveguide which is intrinsic and
contains five 6nm quantum wells which are the gain medium. The quantum wells are what
give you the gain. That's where the carriers collect and that's where the stimulated emission
occurs to give the light output. We have both the electrical confinement and the optical
confinement. With this smaller band gap we have a higher refractive index so this forms
the waveguide too. Now this is a traditional structure which is basically what's used in
the lasers used in optical communications for the transatlantic cables, any other situation.
What we wanted to do was get higher output power so we took some lessons from the high
power community, the high power laser community, and what they've done is interestingly reduced
a number of wells but also they introduced this far field reduction layer which is a
weak waveguide on the n-side of the substrate, of the active layer I mean, so down towards
the substrate. Now what this does is affects the optical field. If we look in the optical
field in this structure you can see that the maximum is in the middle of the quantum well
layer. Here it's being pulled over towards the n-side layer and it does two things. It's
reducing the overlap with the wells but it also reduces the overlap with this P material
and that reduces the free carrier absorption. The free carrier absorption coefficient is
much higher in P-type doping than with N-type doping. The photon hole cross-section, interaction
cross-section is much higher. So what they're doing is they're reducing the losses in the
cavity in order to get more light out. So we wanted more power out of the mode-locked
lasers so we decided to try and use this structure. This is basically a mode-locked laser. We
have, this is the bit of semiconductor and that this long, these are gold contacts and
this long contact is the gain section. This short contact here is the saturable absorber.
That's about 2% of the cavity length. We have a 10 micron isolation gap where we etch the
contact layer between the two contacts, between the gain section and the absorber section
to, so we don't have current leaking into the saturable absorber. And the cavity length
here is 10, is just over one millimeter. And this gives us a mode-locking frequency
of about 40 gigahertz. The saturable absorber itself is about 20, is 25 microns in this
device. And as I said, the output wavelength is 1.55 microns. This is just a picture of
the cleaved facet where you can see we etch ridges and the active region is here just
below the ridge. It's a single lateral mode, so we have a nice round oval mode here and
this is the metal cladding and you can see here the silica passivation layer and there's
none on the top where we inject current through the top of the ridge. Like current characteristics,
this is the output power against the gain current, the current injected into the gain
section. If we forward bias the gain section and the saturable absorber, we just get Fabry-Perot
laser. And this is, these different curves are increasing amount of voltage on the saturable
absorber. We, they will mode-lock with zero volts, but to sweep the carriers out faster
and improve the recovery time of the saturable absorber, you apply a voltage. And as you
can see, as you increase the voltage, the threshold current increases slightly because
you're increasing the losses. It absorbs more light before it saturates and also the slope
efficiency decreases again because of the higher losses. This is the RF spectrum. Basically,
we couple the light onto a fast photodiode and measure the RF signal from that photodiode
and you can see it's operating just about 40 gigahertz. There's no pulses obvious down
here, which would be a sign that it's not mode-locking properly. Now, here we look at
what we call the autocorrelation trace. Here, we split the pulse in two and recombine the
two halves in a non-linear crystal, which generates second harmonic, the second harmonic
of the light. That intensity is proportional to the square of the intensity in the instantaneous
intensity. So, if you've got the pulses separated, you have a much smaller signal than when,
even though, than when they're overlapped. And this is just a function of time and these,
this is just showing the pulses, a train of pulses. Here, we have a single pulse, and
the pulse length for this is about one picosecond. This is an autocorrelation trace, a single
pulse. And that's the optical spectrum, which is a nice, clean spectrum showing plenty of
modes. Now, when we compare the performance of the new structure, the 3-quantum well structure,
with the 5-quantum well structure, just looking at the CW output power as a function of injection
current, you can see we get powers that are at least twice as much coming out, twice the
power coming out before it starts to roll over. This roll over is due to heating, which
is caused by the light being reabsorbed in the cavity and non-radiative recombination,
which should be a small, but it's mainly the reabsorption of light. And you can see that
reducing the overlap with the P-cladding layer has helped that immensely. This is a comparison
of mode locking map. What we have on both graphs is a map of the gain current against
the saturable absorber bias, with the gain current and saturable absorber bias. The red
area is where we see good mode locking. And you can see that with the 3-quantum well device,
this red area is much higher, is much larger. The blue bit down here is where it's not mode
locking. One of the things to note is that if you drive the current high enough, the
device stops mode locking. This is because the saturable absorber is not absorbing enough
of the light that's oscillating within the cavity to cause mode locking. If we look at
the peak pulse power, we can see this is the peak pulse power against gain current. This
is misleading because this is a much finer scale than this one. You can see this only
goes up to 50 milliwatts. This one goes up to 120 milliwatts. So the pulse output power
we're getting is about three times that with the 3-quantum wells compared with the 5-quantum
well. If we look close up at the peak that we find in the RF frequency, the 5-quantum
well device, we find that the width is 36 kilohertz. For the 3-quantum well device,
that's reduced to 25 kilohertz. This is a measure of the pulse stability. The resolution
bandwidth of the equipment was 10 kilohertz. The narrower that is, the more stable the
mode locking frequency is. We also see that when we measured the timing, calculated the
timing jitter between pulses, it's 4.7 picoseconds for the 5-quantum well structure and 1.2 picoseconds
for the 3-quantum well structure. Why is this happening? Well, this here is the material
gain as a function of the sheet carrier density. That's basically the sheet carrier density
is basically how much current you're injecting. If you look here, the one that I particularly
want to look at is this indium gallium arsenide quantum well. You can see that as you increase
the current, the gain increases rapidly to begin with and then it rolls over to less
rapid. Most 5-quantum well structures are operating here. Our structure's operating
up here, where the gain change for the change in carrier density is much smaller. Now, the
gain is related to the refractive index through the Cramer's-Kronig relationship. If you have
less change in gain, you have less change in refractive index. We have a dynamic relationship
for equilibrium in that the carrier density is changing slightly. Even though we're injecting
a constant current through the cavity, the carrier density is changing slightly. When
the pulse is propagating around in the cavity, the carrier density and the refractive index
is changing as it goes through. But if we operate up here, the change in refractive
index is smaller. The change in refractive index causes a change in speed. That means
that the change in speed from pulse to pulse is smaller if the change in refractive index
from pulse to pulse is smaller. That leads to the more stable mode locking and the lower
timing jitter. Another thing that we looked at was to introduce a passive section into
the cavity. We did this by quantum well intermixing. Basically, you take these nice quantum wells
that the growers spend all the time perfecting and then pattern the top so that you have
a cap that will cause point defects to diffuse through them and get the wells to intermix
with the barriers. That means that why we do that is to increase the band gap. The band
gap here is larger than the band gap here, so the light generated here is not absorbed
here even though we're not injecting current. This again is just over one millimetre long,
so we have comparable mode locking frequency. If we look at the light current characteristic,
this is a Fabry-Perot laser without the passive cavity. Here we have one, the mode locking
laser with the passive cavity. We have a distinct increase in threshold current due to increase
in losses because of the satural absorber and there are some losses in that passive
section, but they are relatively small. The power output is also a bit lower. If we look
at the mode locking map again, as we showed before, the gain current against the saturable
absorber bias, you can see in this case we have a much larger area where it's mode locking.
The other thing that we found was the mode locking pulses, the pulse width is very even
throughout this area. This is for a cavity without the passive section. You can see that
as you increase the current, the pulse width goes up, the pulse length goes up. Here it
remains very flat. This is due to the fact that we have reduced what's called the self-phase
modulation. As the pulse goes through the cavity, the front of the pulse sees a slightly
different refractive index from the back because it's caused stimulated emission and
reduced the number of carriers as it goes through. The back of the pulse has a slightly
different refractive index. In the passive section, this doesn't happen because we're
not taking out any carriers. We've reduced the self-phase modulation. That self-phase
modulation causes the change in refractive index changes the speed of the light as the
pulse goes through the cavity, within the pulse. It's changing the speed of the light
within the pulse and spreads it out. By removing some of that self-phase modulation, we've
managed to narrow the pulses and it's meant that increasing the current has had a smaller
effect on spreading out the current pulse. When we actually measured the current pulse,
we find that the pulse length is about 490 femtoseconds. By changing the design of the
epilayer, we've managed to improve the timing jitter from these devices. By introducing
a passive section in the cavity, we can reduce the pulse length. Thank you.
Time for a few questions. What's the highest power output that you can get from a...
The highest power we've seen out is about 60 milliwatts, I believe. No, maybe 30 milliwatts.
The traditional, the structure that I showed you to begin with, you only get a few milliwatts
out. We managed to get that up to about 30 or 40 milliwatts. We have integrated amplifiers
and managed to get out much higher powers coming up to a watt and several watts peak
power. Oh, these are broad because we need several modes. The more modes you can lock
together, the narrower the pulse. The individual? We've not looked at that. We've not been interested
in that. Let's thank all our speakers again. We have
about 15 minutes for a break and then please...

